
\documentclass[11pt,halfline,a4paper,]{ouparticle}

% Packages I think are necessary for basic Rmarkdown functionality
\usepackage{hyperref}
\usepackage{graphicx}
\usepackage{listings}
\usepackage{xcolor}
\usepackage{fancyvrb}
\usepackage{framed}

% Link coloring
\hypersetup{breaklinks=true,
            bookmarks=true,
            pdfauthor={},
            pdftitle={Supporting Information}
            }


%% To allow better options for figure placement
%\usepackage{float}

% Packages that are supposedly required by OUP sty file
\usepackage{amssymb, amsmath, geometry, amsfonts, verbatim, endnotes, setspace}

% use upquote if available, for straight quotes in verbatim environments
\IfFileExists{upquote.sty}{\usepackage{upquote}}{}

% Macros for dealing with affiliations, footnotes, etc.
\makeatletter
\def\Newlabel#1#2#3{\expandafter\gdef\csname #1@#2\endcsname{#3}}

\def\Ref#1#2{\@ifundefined{#1@#2}{???}{\csname #1@#2\endcsname}}

\newcommand*\samethanks[1][\value{footnote}]{\footnotemark[#1]}

\newcommand*\ifcounter[1]{%
  \ifcsname c@#1\endcsname
    \expandafter\@firstoftwo
  \else
    \expandafter\@secondoftwo
  \fi
}

\newcommand*\thanksbycode[1]{%
  \ifcounter{FNCT@#1}
    {\samethanks[\value{FNCT@#1}]}
    {\thanks{\Ref{FN}{#1}}\newcounter{FNCT@#1}\setcounter{FNCT@#1}{\value{footnote}}}
}

% Create labels for Addresses if the are given in Elsevier format

% Create labels for Footnotes if the are given in Elsevier format

% Part for setting citation format package: natbib

% Part for setting citation format package: biblatex


% tightlist command for lists without linebreak
\providecommand{\tightlist}{%
  \setlength{\itemsep}{0pt}\setlength{\parskip}{0pt}}



\usepackage{booktabs}
\usepackage{longtable}
\usepackage{array}
\usepackage{multirow}
\usepackage{wrapfig}
\usepackage{float}
\usepackage{colortbl}
\usepackage{pdflscape}
\usepackage{tabu}
\usepackage{threeparttable}
\usepackage{threeparttablex}
\usepackage[normalem]{ulem}
\usepackage{makecell}
\usepackage{xcolor}

\begin{document}

\title{Supporting Information}

\author{%
%
% Code for old style authors field
%
% Add \and if both authors and author
%
%
% Code for new (elsevier) style author field
%
}

\abstract{}

\date{\today}

\keywords{}

\maketitle



\hypertarget{irt-model-details-and-diagnostics}{%
\section{IRT model details and
diagnostics}\label{irt-model-details-and-diagnostics}}

In analysing our data, we used scales tapping into the following in our
analysis:

\begin{enumerate}
\def\labelenumi{\arabic{enumi}.}
\tightlist
\item
  Subjective numeracy (independent variable)
\item
  Trust in government (independent variable)
\item
  Knowledge about COVID-19 (independent variable)
\item
  Contact with COVID-19 (independent variable)
\item
  Support for restrictions (dependent variable)
\item
  Willingness to engage in activities (dependent variable)
\item
  Level of concern about COVID-19 (dependent variable)
\item
  Perceptions of government performance (dependent variable)
\end{enumerate}

2-8 were constructed out of ordinal survey items using item response
theory (IRT) modelling. IRT models are used to measure latent traits
assumed to fall on a continuous scale. Values on that scale are usually
referred to by way of the Greek letter \(\theta\) (theta), and taken to
range from -\(\infty\) to +\(\infty\), with a mean of 0 and standard
deviation of 1. This means that, while the individual \(\theta\) values
ascribed to any particular respondent has no intrinsic meaning, it can
nevertheless be interpreted relative to an estimated population mean.
Subjective numeracy was modelled using the mean response across three
established items (more on this below), as per standard practice.
Further details are provided below.

A valid IRT model should meet three conditions:

\begin{itemize}
\tightlist
\item
  unidimensionality (the scale taps into only one dimension/trait),
\item
  local independence (items are uncorrelated after conditioning on the
  measured trait), and
\item
  good model fit.
\end{itemize}

We evaluated unidimensionality using parallel analysis (where finding no
more than 1 factor indicates likely unidimensionality), and local
independence using Yen's (1993) Q3. Yen suggests a screening value of
+/-0.2, such that, anything larger than that, is evidence of local
dependence. However, as pointed out by de Ayala (2009: 137), that
screening value is most appropriate for long scales (35 items or more),
as shorter scales like ours can be expected to give higher values. We
evaluated model fit by inspecting empirical plots.

In the case of the scales that had not been pre-validated through
previous work (i.e., 3-8), we did not know in advance exactly what items
would end up making for a valid IRT model. For each scale, only the
subset of items that made for a valid IRT models were included, with
details provided in subsequent sections. Doing otherwise would risk
basing our analysis on flawed models potentially giving us inaccurate
estimates.

\hypertarget{subjective-numeracy}{%
\subsection{Subjective numeracy}\label{subjective-numeracy}}

Subjective numeracy was measured with reference to the mean response to
the following three items (McNaughton et al.~2015), with participants
being asked to indicate on a scale from 1 to 6 whether they considered
themselves `Not good at all' (1) or `Extremely good' (6):

\begin{enumerate}
\def\labelenumi{\arabic{enumi}.}
\tightlist
\item
  How good are you at working with fractions
\item
  How good are you at figuring out how much a shirt will cost if it is
  25\% off?
\item
  How often do you find numerical information to be useful?
\end{enumerate}

\hypertarget{trust}{%
\subsection{Trust}\label{trust}}

Our trust scale was built from the following items from Devine et
al.~(2020), with response options ranging from `strongly disagree' (1)
to `strongly agree' (5):

\begin{enumerate}
\def\labelenumi{\arabic{enumi}.}
\tightlist
\item
  The government understands the needs of my community.
\item
  The government usually has good intentions.
\item
  In general, the government usually does the right thing.
\end{enumerate}

Since non-binary response options, we here used a generalized partial
credit model, with the R package mirt (Chalmers 2012). The items had
good discrimination values (2.116, 2.803, and 4.862), loaded well onto
the factor, were well-ordered, and made for a unidimensional scale. By
way of scale reliability, the scale discriminated best at around the
mean (an approximate information value of 10 in the \(\theta\) range of
-1.75 to 2). The Q3 values were acceptable (a maximum of -0.591, for
items 2 and 3), given the length of the scale.

\hypertarget{knowledge}{%
\subsection{Knowledge}\label{knowledge}}

We included the following seven COVID-19 knowledge items in our survey,
in each case asking respondents to indicate whether the statement was
true or false:

\begin{enumerate}
\def\labelenumi{\arabic{enumi}.}
\tightlist
\item
  COVID-19 is a virus (false)
\item
  COVID-19 is caused by bacteria (false)
\item
  COVID-19 can be transmitted in areas with hot and humid climate (true)
\item
  There is currently no vaccine to protect against COVID-19 (true)
\item
  Most people who get COVID-19 recover from it (true)
\item
  COVID-19 can be transmitted through mosquito bites (false)
\item
  Antibiotics are effective in preventing and treating COVID-19 (false)
\end{enumerate}

Using the R package mirt (Chalmers 2012) to fit a two-parameter model
(2PL), we determined that items 3, 4, and 5 made for a unidimensional
scale with acceptable discrimination values (1.240, 1.372, and 1.337,
respectively). By way of scale reliability, the test information
function showed the scale to discriminate best below the mean (an
approximate information of 1 at a \(\theta\) of -2). The maximum Q3
value was acceptable (-0.192 between items 3 and 5), as was the model
fit.

\hypertarget{contact}{%
\subsection{Contact}\label{contact}}

To measure participants' contact with COVID-19, we also included the
following items in the survey, asking participants to designate whether
the statement was true or false in their case:

\begin{enumerate}
\def\labelenumi{\arabic{enumi}.}
\tightlist
\item
  I have tested positive for coronavirus or for coronavirus antibodies
\item
  I have not been tested, but am highly confident that I either have or
  have had coronavirus
\item
  Someone I know has tested positive for coronavirus or for coronavirus
  antibodies
\item
  Someone I know has passed away from a confirmed case of coronavirus
\item
  My job involves treating coronavirus patients, or supporting those who
  are giving treatment
\end{enumerate}

Using mirt, items 4, 5, and 6 were found to form a unidimensional scale
with good discrimination values (1.473, 1.326, and 1.321). By way of
scale reliability, the scale discriminated best at and above the mean
(an approximate information of 0.9 at a \(\theta\) of 1). The maximum Q3
value (-0.239, between items 3 and 4) was acceptable, as was the model
fit.

\hypertarget{support-for-restrictions}{%
\subsection{Support for restrictions}\label{support-for-restrictions}}

In measuring support for restrictions, we used the following statements,
where participants were asked to tick each item corresponding to a
restriction they support:

\begin{enumerate}
\def\labelenumi{\arabic{enumi}.}
\tightlist
\item
  Quarantining anyone who has been in contact with a contaminated
  patient
\item
  Temporarily closing schools
\item
  Cancelling large events
\item
  Cancelling routine hospital procedures
\item
  Stopping all inbound international flights from countries with
  confirmed cases of coronavirus
\item
  Quarantining all inbound international flights from countries with
  confirmed cases of coronavirus
\item
  Encouraging social distancing of 1 metre
\item
  Requiring wearing masks on public transport
\item
  Requiring wearing masks in all indoor public places
\end{enumerate}

Using mirt, three of the items---3, 7, and 9---were found to have good
discrimination values (1.861, 2.400, and 2.112) and made for a
unidimensional scale. By way of scale reliability, the scale
discriminated best below the mean (an approximate information value of 3
at a \(\theta\) value of -1.75), and the maximum Q3 value (-0.277
between 7 and 9) was acceptable since the scale was short, as discussed
above.

\hypertarget{willingness-to-engage-in-activities}{%
\subsection{Willingness to engage in
activities}\label{willingness-to-engage-in-activities}}

To measure participants' willingness to engage in activities, we used
the following items, where participants were asked to tick each item
corresponding to an activity they expected to engage in over the coming
month:

\begin{enumerate}
\def\labelenumi{\arabic{enumi}.}
\tightlist
\item
  Going to the hairdressers
\item
  Visiting an indoor museum or exhibition
\item
  Going to bars and restaurants
\item
  Using public toilets
\item
  Using public transport
\item
  Going to indoor cinemas or theatres
\item
  Taking holidays abroad
\item
  Attending places of worship
\item
  Going to indoor gyms/leisure centres/swimming pools
\item
  Going to large public gatherings (sport/music events)
\end{enumerate}

Using a 2PL model in mirt, three of the items (3, 6, and 10) were
determined to have good discrimination values (3.452, 1.714, and 2.260)
and made for a unidimensional scale. By way of scale reliability, the
scale discriminated best around the mean (an information value of about
3.25 for a \(\theta\) of 0, with an additional lower peak of around 1.75
for a \(\theta\) of 2), and the maximum Q3 value (0.409 between 3 and 6)
was acceptable.

\hypertarget{level-of-concern}{%
\subsection{Level of concern}\label{level-of-concern}}

For level of concern, we presented participants with seven statements,
asking them to use a scale from 1-4 to indicate whether they were `not
worried at all' (1) or `extremely worried' (4) of the following, with
wordings taken from YouGov (2020):

\begin{enumerate}
\def\labelenumi{\arabic{enumi}.}
\tightlist
\item
  Contracting COVID-19 (the disease caused by coronavirus)
\item
  Becoming seriously unwell or dying
\item
  Friends or family becoming seriously unwell or dying
\item
  Finances being severely affected
\item
  Losing your job
\item
  Your children's education suffering
\item
  There being a long-lasting negative impact on society
\end{enumerate}

Using a generalized partial credit model in mirt, we found that three of
the items (1, 2, and 3) loaded well onto the factor, had good
discrimination values (3.740, 4.011, and 1.477), were well-ordered, and
made for a scale that was likely unidimensional. By way of scale
reliability, the scale discriminated best in the \(\theta\) range of -1
to 1 (an approximate information value of 9). The Q3 values was
acceptable (a maximum of -0.723, for items 1 and 2), given the length of
the scale.

\hypertarget{perceptions-of-government-performance}{%
\subsection{Perceptions of government
performance}\label{perceptions-of-government-performance}}

To measure perceptions of government performance, we asked participants
to use a scale of 1-5 to indicate whether they `strongly disagree' (1)
or `strongly agree' (5) with the following statements, borrowed from
Ipsos MORI (2020):

\begin{enumerate}
\def\labelenumi{\arabic{enumi}.}
\tightlist
\item
  I have found the communication and advice from the UK government
  helpful.
\item
  The UK government's advice on how to protect myself and others has
  been effective.
\item
  The UK government's response to the coronavirus has been clear and
  consistent.
\item
  The UK government's plan has adapted well to the changing scientific
  information and situation.
\item
  Compared with other countries, the UK government has responded well to
  the coronavirus outbreak.
\item
  The UK government has done a good job of protecting UK residents
  through its response to the coronavirus.
\end{enumerate}

Using mirt, we found that three of the items (4, 5, and 6) loaded well,
had good discrimination values (2.314, 2.689, and 5.590), were
well-ordered, and made for a unidimensional scale. By way of scale
reliability, the scale discriminated best in the \(\theta\) range of -1
to 2, with an approximate maximum information value of 16 around 0.5.
The maximum Q3 value was acceptable (0.589 between items 4 an 6), given
the length of the scale.

\hypertarget{manipulation-check}{%
\section{Manipulation check}\label{manipulation-check}}

We included a post-treatment manipulation check, asking respondents
assigned to either treatment group to indicate whether the image they
had previously seen contained ``one'' or ``more than one'' country. Fig.
\ref{fig:fig1} displays the breakdown by condition. It shows that the
manipulation worked as intended: respondents in the UK-only condition
overwhelmingly reported seeing one country in the image, while those in
the comparative condition reported seeing more than one country. In the
main analysis, we excluded respondents who gave a response that was
incorrect for their assigned condition. However, we also re-ran the
analyses using the full sample (see note 10 in the main body).

\begin{figure}[h]
\includegraphics[width=1\linewidth]{covid_appendix_files/figure-latex/fig1-1} \caption{Distribution of responses to the manipulation check by treatment condition}\label{fig:fig1}
\end{figure}

\hypertarget{sample-details}{%
\section{Sample details}\label{sample-details}}

Table \ref{tab:tab1a} provides details on the distribution of
participants across the three conditions along demographic variables,
and Table \ref{tab:tab1b} does the same for for pre-treatment variables.
The BES benchmark proportions in the former table are from
\href{https://www.britishelectionstudy.com/data-object/wave-20-of-the-2014-2023-british-election-study-internet-panel/}{Wave
20 of the 2014-2023 British Election Study Internet Panel}, fielded in
June 2020 (N = 31,468), so a month before our survey data was collected.
Table \ref{tab:tab2} gives details on the achieved proportion and sample
targets by gender and ethnicity.

\begin{table}

\caption{\label{tab:tab1}Demographic variables by treatment with BES benchmarks (Wave 20 of the BES Internet Panel, June 2020; survey weights applied)}
\centering
\begin{tabular}[t]{lrrrrr}
\toprule
  & Control & \makecell[c]{UK-Only\\Treatment} & \makecell[c]{Comparative\\Treatment} & SD & \makecell[c]{BES\\Benchmark}\\
\midrule
Age (mean): & 36.0 & 35.0 & 35.0 & 0.58 & 48.3\\
Female (perc.) & 50.9 & 50.4 & 51.4 & 0.50 & 51.5\\
Male (perc.) & 49.1 & 49.6 & 48.6 & 0.50 & 48.5\\
Educ: GCSE or less (perc.) & 11.6 & 11.0 & 13.8 & 1.47 & 26.0\\
Educ: A-level (perc.) & 31.9 & 31.4 & 30.2 & 0.87 & 28.7\\
Educ: Undergraduate (perc.) & 41.3 & 41.0 & 41.6 & 0.30 & 36.6\\
Educ: Postgraduate (perc.) & 15.3 & 16.6 & 14.5 & 1.06 & 8.8\\
Ethnicity: White (perc.) & 86.2 & 86.3 & 85.9 & 0.21 & 91.2\\
Ethnicity: Other (perc.) & 13.8 & 13.7 & 14.1 & 0.21 & 8.8\\
Income: above mean (30k) (perc.) & 56.4 & 54.6 & 54.9 & 0.96 & 51.0\\
Party: Conservatives (perc.) & 16.3 & 18.0 & 21.9 & 2.87 & 30.4\\
Party: Labour (perc.) & 40.3 & 39.1 & 37.3 & 1.51 & 25.6\\
Party: Lib Dem (perc.) & 10.8 & 9.3 & 8.5 & 1.17 & 6.6\\
Party: Green Party (perc.) & 8.0 & 9.8 & 8.4 & 0.95 & 3.3\\
Party: Other party (perc.) & 8.2 & 5.3 & 6.7 & 1.45 & 6.2\\
Party: No party (perc.) & 16.5 & 18.5 & 17.1 & 1.03 & 27.9\\
EU vote: Leave (perc.) & 23.9 & 23.7 & 24.4 & 0.36 & 51.8\\
EU vote: Remain (perc.) & 65.7 & 64.8 & 63.7 & 1.00 & 48.0\\
EU vote: Didn't vote (perc.) & 10.4 & 11.5 & 11.9 & 0.78 & 0.2\\
N & 1001.0 & 921.0 & 995.0 & 44.56 & 31468.0\\
\bottomrule
\end{tabular}
\end{table}

\begin{table}

\caption{\label{tab:tab2}Pre-treatment variables by treatment}
\centering
\begin{tabular}[t]{lrrrr}
\toprule
  & Control & \makecell[c]{UK-Only\\Treatment} & \makecell[c]{Comparative\\Treatment} & SD\\
\midrule
Contact (mean) & -0.01 & 0.02 & 0.00 & 0.02\\
Knowledge (mean) & 0.00 & 0.01 & -0.01 & 0.01\\
Trust (mean) & 0.00 & -0.04 & 0.04 & 0.04\\
Numeracy (mean) & 4.14 & 4.22 & 4.12 & 0.05\\
N & 1001.00 & 921.00 & 995.00 & 0.87\\
\bottomrule
\end{tabular}
\end{table}

\begin{table}

\caption{\label{tab:tab3}Achieved sample proportions and targets by gender and ethnicity}
\centering
\begin{tabular}[t]{lcc}
\toprule
Ethnicity & Female & Male\\
\midrule
White & 44.1 & 42.1\\
\textit{   Target} & \textit{44.0} & \textit{42.0}\\
Other & 6.8 & 7.0\\
\textit{   Target} & \textit{7.0} & \textit{7.0}\\
\bottomrule
\end{tabular}
\end{table}

\hypertarget{model-details}{%
\section{Model details}\label{model-details}}

Tables 3-7 provide details on the models used for the analysis.
Inspection of diagnostic plots suggested that several of the models
exhibited minor problems with non-normally distributed residuals and
some heteroscedasticity. Robust standard errors were therefore used,
computed with R's \texttt{sandwich} package (Zeileis 2004). The three
models involving the dependent variable `Support for restrictions' also
exhibited some non-linearity. For that reason, a binary version of that
variable was created, with values above the mean coded as 1, and all
others as 0, and a corresponding logistic model was fitted for each of
these three linear models. These can be found in Table 8. These models
do not change the analysis provided in relation to the linear models in
the main text.

\begin{table}[h!]
\caption{Main effects (OLS with robust standard errors)}
\begin{center}
\begin{tabular}{l c c c c}
\hline
 & Gov. Performance & Activity & Concern & Restrictions \\
\hline
(Intercept)      & $0.422 \; (0.075)^{***}$    & $0.471 \; (0.083)^{***}$    & $0.060 \; (0.104)$          & $-0.191 \; (0.071)^{**}$   \\
Cond.: UK-Only   & $0.057 \; (0.030)^{\cdot}$  & $-0.041 \; (0.033)$         & $0.035 \; (0.040)$          & $0.042 \; (0.028)$         \\
Cond.: Comp.     & $-0.037 \; (0.029)$         & $-0.050 \; (0.033)$         & $0.030 \; (0.041)$          & $0.007 \; (0.029)$         \\
Contact          & $-0.039 \; (0.022)^{\cdot}$ & $0.089 \; (0.024)^{***}$    & $0.090 \; (0.029)^{**}$     & $0.032 \; (0.019)^{\cdot}$ \\
Knowledge        & $-0.107 \; (0.023)^{***}$   & $0.028 \; (0.024)$          & $-0.087 \; (0.031)^{**}$    & $0.119 \; (0.022)^{***}$   \\
Numeracy         & $-0.019 \; (0.010)^{\cdot}$ & $0.022 \; (0.011)^{*}$      & $-0.063 \; (0.015)^{***}$   & $0.003 \; (0.010)$         \\
Trust            & $0.624 \; (0.016)^{***}$    & $0.060 \; (0.017)^{***}$    & $-0.015 \; (0.021)$         & $0.019 \; (0.015)$         \\
Age              & $-0.001 \; (0.001)$         & $-0.012 \; (0.001)^{***}$   & $0.008 \; (0.001)^{***}$    & $-0.000 \; (0.001)$        \\
Male             & $-0.021 \; (0.026)$         & $0.031 \; (0.029)$          & $-0.250 \; (0.036)^{***}$   & $-0.080 \; (0.025)^{**}$   \\
GCSE or below    & $0.041 \; (0.042)$          & $-0.080 \; (0.047)^{\cdot}$ & $-0.022 \; (0.061)$         & $0.019 \; (0.044)$         \\
Undergraduate    & $-0.056 \; (0.030)^{\cdot}$ & $0.028 \; (0.033)$          & $-0.032 \; (0.041)$         & $0.014 \; (0.028)$         \\
Postgraduate     & $-0.083 \; (0.040)^{*}$     & $0.050 \; (0.044)$          & $-0.081 \; (0.054)$         & $0.007 \; (0.036)$         \\
Other ethnicity  & $0.062 \; (0.038)^{\cdot}$  & $-0.175 \; (0.041)^{***}$   & $0.103 \; (0.051)^{*}$      & $-0.006 \; (0.034)$        \\
Income $>$ 30k   & $0.026 \; (0.025)$          & $0.047 \; (0.027)^{\cdot}$  & $-0.060 \; (0.034)^{\cdot}$ & $-0.005 \; (0.024)$        \\
No party         & $-0.110 \; (0.046)^{*}$     & $-0.122 \; (0.050)^{*}$     & $-0.062 \; (0.064)$         & $0.007 \; (0.047)$         \\
Lib Dem          & $-0.282 \; (0.056)^{***}$   & $-0.170 \; (0.057)^{**}$    & $0.130 \; (0.071)^{\cdot}$  & $0.161 \; (0.048)^{***}$   \\
Labour           & $-0.247 \; (0.043)^{***}$   & $-0.112 \; (0.046)^{*}$     & $0.202 \; (0.057)^{***}$    & $0.095 \; (0.042)^{*}$     \\
Green            & $-0.196 \; (0.057)^{***}$   & $-0.223 \; (0.062)^{***}$   & $0.080 \; (0.078)$          & $0.106 \; (0.055)^{\cdot}$ \\
Other party      & $-0.353 \; (0.062)^{***}$   & $-0.194 \; (0.062)^{**}$    & $0.082 \; (0.079)$          & $0.115 \; (0.059)^{*}$     \\
EU: Did not vote & $-0.066 \; (0.051)$         & $-0.007 \; (0.054)$         & $-0.010 \; (0.068)$         & $0.057 \; (0.051)$         \\
EU: Remain       & $-0.161 \; (0.034)^{***}$   & $-0.036 \; (0.038)$         & $-0.026 \; (0.047)$         & $0.186 \; (0.035)^{***}$   \\
\hline
R$^2$            & $0.509$                     & $0.071$                     & $0.068$                     & $0.050$                    \\
Adj. R$^2$       & $0.506$                     & $0.064$                     & $0.062$                     & $0.044$                    \\
Num. obs.        & $2917$                      & $2917$                      & $2917$                      & $2917$                     \\
\hline
\multicolumn{5}{l}{\scriptsize{$^{***}p<0.001$; $^{**}p<0.01$; $^{*}p<0.05$; $^{\cdot}p<0.1$}}
\end{tabular}
\label{tab4}
\end{center}
\end{table}

\begin{table}[h!]
\caption{Interaction effects: Condition x Trust (OLS with robust standard errors)}
\begin{center}
\begin{tabular}{l c}
\hline
 & Gov. Performance \\
\hline
(Intercept)        & $0.422 \; (0.075)^{***}$    \\
Cond.: UK-Only     & $0.056 \; (0.030)^{\cdot}$  \\
x Trust            & $-0.108 \; (0.031)^{***}$   \\
Cond.: Comparative & $-0.037 \; (0.029)$         \\
x Trust            & $-0.078 \; (0.031)^{*}$     \\
Contact            & $-0.041 \; (0.022)^{\cdot}$ \\
Knowledge          & $-0.109 \; (0.023)^{***}$   \\
Numeracy           & $-0.019 \; (0.010)^{\cdot}$ \\
Trust              & $0.687 \; (0.022)^{***}$    \\
Age                & $-0.001 \; (0.001)$         \\
Male               & $-0.021 \; (0.026)$         \\
GCSE or below      & $0.040 \; (0.042)$          \\
Undergraduate      & $-0.055 \; (0.030)^{\cdot}$ \\
Postgraduate       & $-0.082 \; (0.040)^{*}$     \\
Other ethnicity    & $0.063 \; (0.038)^{\cdot}$  \\
Income $>$ 30k     & $0.025 \; (0.025)$          \\
No party           & $-0.117 \; (0.046)^{*}$     \\
Lib Dem            & $-0.281 \; (0.056)^{***}$   \\
Labour             & $-0.247 \; (0.043)^{***}$   \\
Green              & $-0.201 \; (0.057)^{***}$   \\
Other party        & $-0.351 \; (0.061)^{***}$   \\
EU: Did not vote   & $-0.063 \; (0.051)$         \\
EU: remain         & $-0.160 \; (0.034)^{***}$   \\
\hline
R$^2$              & $0.511$                     \\
Adj. R$^2$         & $0.507$                     \\
Num. obs.          & $2917$                      \\
\hline
\multicolumn{2}{l}{\scriptsize{$^{***}p<0.001$; $^{**}p<0.01$; $^{*}p<0.05$; $^{\cdot}p<0.1$}}
\end{tabular}
\label{tab5}
\end{center}
\end{table}

\begin{table}[h!]
\caption{Interaction effects: Partisanship (OLS with robust standard errors)}
\begin{center}
\begin{tabular}{l c}
\hline
 & Gov. Performance \\
\hline
(Intercept)        & $0.446 \; (0.079)^{***}$    \\
Cond.: UK-Only     & $0.016 \; (0.067)$          \\
x Green Party      & $0.001 \; (0.125)$          \\
x Labour           & $0.076 \; (0.082)$          \\
x Lib Dem          & $-0.097 \; (0.127)$         \\
x No party         & $0.064 \; (0.098)$          \\
x Other party      & $0.161 \; (0.152)$          \\
Cond.: Comparative & $-0.066 \; (0.066)$         \\
x Green Party      & $-0.001 \; (0.125)$         \\
x Labour           & $0.042 \; (0.080)$          \\
x Lib Dem          & $-0.107 \; (0.121)$         \\
x No party         & $0.111 \; (0.097)$          \\
x Other party      & $0.040 \; (0.129)$          \\
Contact            & $-0.040 \; (0.022)^{\cdot}$ \\
Knowledge          & $-0.105 \; (0.023)^{***}$   \\
Numeracy           & $-0.018 \; (0.010)^{\cdot}$ \\
Trust              & $0.624 \; (0.016)^{***}$    \\
Age                & $-0.001 \; (0.001)$         \\
Male               & $-0.021 \; (0.026)$         \\
GCSE or below      & $0.040 \; (0.042)$          \\
Undergraduate      & $-0.057 \; (0.030)^{\cdot}$ \\
Postgraduate       & $-0.085 \; (0.040)^{*}$     \\
Other ethnicity    & $0.062 \; (0.038)$          \\
Income $>$ 30k     & $0.024 \; (0.025)$          \\
No party           & $-0.169 \; (0.066)^{*}$     \\
Lib Dem            & $-0.222 \; (0.084)^{**}$    \\
Labour             & $-0.286 \; (0.059)^{***}$   \\
Green party        & $-0.195 \; (0.098)^{*}$     \\
Other party        & $-0.410 \; (0.093)^{***}$   \\
EU: Did not vote   & $-0.066 \; (0.051)$         \\
EU: Remain         & $-0.161 \; (0.034)^{***}$   \\
\hline
R$^2$              & $0.510$                     \\
Adj. R$^2$         & $0.505$                     \\
Num. obs.          & $2917$                      \\
\hline
\multicolumn{2}{l}{\scriptsize{$^{***}p<0.001$; $^{**}p<0.01$; $^{*}p<0.05$; $^{\cdot}p<0.1$}}
\end{tabular}
\label{tab6}
\end{center}
\end{table}

\begin{table}[h!]
\caption{Interaction effects: Condition x Contact (OLS with robust standard errors)}
\begin{center}
\begin{tabular}{l c c c c}
\hline
 & Gov. Performance & Activity & Concern & Restrictions \\
\hline
(Intercept)        & $0.422 \; (0.075)^{***}$    & $0.471 \; (0.083)^{***}$    & $0.059 \; (0.103)$          & $-0.191 \; (0.071)^{**}$   \\
Cond.: UK-Only     & $0.057 \; (0.030)^{\cdot}$  & $-0.041 \; (0.033)$         & $0.035 \; (0.040)$          & $0.042 \; (0.028)$         \\
x Contact          & $-0.034 \; (0.053)$         & $0.052 \; (0.058)$          & $0.052 \; (0.068)$          & $-0.038 \; (0.046)$        \\
Cond.: Comparative & $-0.037 \; (0.029)$         & $-0.050 \; (0.033)$         & $0.031 \; (0.041)$          & $0.006 \; (0.029)$         \\
x Contact          & $-0.031 \; (0.051)$         & $0.028 \; (0.057)$          & $0.096 \; (0.070)$          & $-0.045 \; (0.046)$        \\
Contact            & $-0.017 \; (0.036)$         & $0.063 \; (0.041)$          & $0.039 \; (0.049)$          & $0.060 \; (0.031)^{\cdot}$ \\
Knowledge          & $-0.107 \; (0.023)^{***}$   & $0.028 \; (0.024)$          & $-0.086 \; (0.031)^{**}$    & $0.119 \; (0.023)^{***}$   \\
Numeracy           & $-0.019 \; (0.010)^{\cdot}$ & $0.023 \; (0.011)^{*}$      & $-0.063 \; (0.015)^{***}$   & $0.003 \; (0.010)$         \\
Trust              & $0.624 \; (0.016)^{***}$    & $0.060 \; (0.017)^{***}$    & $-0.015 \; (0.021)$         & $0.019 \; (0.015)$         \\
Age                & $-0.001 \; (0.001)$         & $-0.012 \; (0.001)^{***}$   & $0.008 \; (0.001)^{***}$    & $-0.000 \; (0.001)$        \\
Male               & $-0.021 \; (0.026)$         & $0.032 \; (0.029)$          & $-0.250 \; (0.036)^{***}$   & $-0.080 \; (0.025)^{**}$   \\
GCSE or below      & $0.040 \; (0.042)$          & $-0.078 \; (0.047)^{\cdot}$ & $-0.020 \; (0.061)$         & $0.018 \; (0.044)$         \\
Undergraduate      & $-0.056 \; (0.030)^{\cdot}$ & $0.029 \; (0.033)$          & $-0.029 \; (0.041)$         & $0.012 \; (0.028)$         \\
Postgraduate       & $-0.084 \; (0.040)^{*}$     & $0.051 \; (0.044)$          & $-0.080 \; (0.054)$         & $0.007 \; (0.036)$         \\
Other ethnicity    & $0.063 \; (0.038)^{\cdot}$  & $-0.176 \; (0.041)^{***}$   & $0.103 \; (0.051)^{*}$      & $-0.005 \; (0.034)$        \\
Income $>$ 30k     & $0.026 \; (0.025)$          & $0.047 \; (0.027)^{\cdot}$  & $-0.060 \; (0.034)^{\cdot}$ & $-0.005 \; (0.024)$        \\
No party           & $-0.110 \; (0.046)^{*}$     & $-0.123 \; (0.050)^{*}$     & $-0.063 \; (0.064)$         & $0.007 \; (0.047)$         \\
Lib Dem            & $-0.282 \; (0.056)^{***}$   & $-0.170 \; (0.057)^{**}$    & $0.129 \; (0.071)^{\cdot}$  & $0.161 \; (0.048)^{***}$   \\
Labour             & $-0.246 \; (0.043)^{***}$   & $-0.113 \; (0.046)^{*}$     & $0.201 \; (0.057)^{***}$    & $0.096 \; (0.042)^{*}$     \\
Green Party        & $-0.195 \; (0.057)^{***}$   & $-0.223 \; (0.062)^{***}$   & $0.077 \; (0.078)$          & $0.107 \; (0.055)^{\cdot}$ \\
Other party        & $-0.351 \; (0.062)^{***}$   & $-0.197 \; (0.062)^{**}$    & $0.076 \; (0.079)$          & $0.118 \; (0.059)^{*}$     \\
EU: Did not vote   & $-0.066 \; (0.051)$         & $-0.007 \; (0.054)$         & $-0.011 \; (0.068)$         & $0.057 \; (0.051)$         \\
EU: Remain         & $-0.161 \; (0.034)^{***}$   & $-0.036 \; (0.038)$         & $-0.026 \; (0.047)$         & $0.186 \; (0.035)^{***}$   \\
\hline
R$^2$              & $0.509$                     & $0.071$                     & $0.069$                     & $0.051$                    \\
Adj. R$^2$         & $0.505$                     & $0.064$                     & $0.062$                     & $0.043$                    \\
Num. obs.          & $2917$                      & $2917$                      & $2917$                      & $2917$                     \\
\hline
\multicolumn{5}{l}{\scriptsize{$^{***}p<0.001$; $^{**}p<0.01$; $^{*}p<0.05$; $^{\cdot}p<0.1$}}
\end{tabular}
\label{tab7}
\end{center}
\end{table}

\begin{table}[h!]
\caption{Interaction effects: Condition x Knowledge (OLS with robust standard errors)}
\begin{center}
\begin{tabular}{l c c c c}
\hline
 & Gov. Performance & Activity & Concern & Restrictions \\
\hline
(Intercept)        & $0.422 \; (0.075)^{***}$    & $0.472 \; (0.083)^{***}$    & $0.060 \; (0.104)$          & $-0.191 \; (0.071)^{**}$   \\
Cond.: UK-Only     & $0.057 \; (0.030)^{\cdot}$  & $-0.041 \; (0.033)$         & $0.035 \; (0.040)$          & $0.042 \; (0.028)$         \\
x Knowledge        & $-0.000 \; (0.054)$         & $0.088 \; (0.057)$          & $0.035 \; (0.072)$          & $-0.000 \; (0.053)$        \\
Cond.: Comparative & $-0.037 \; (0.029)$         & $-0.050 \; (0.033)$         & $0.030 \; (0.041)$          & $0.007 \; (0.028)$         \\
x Knowledge        & $-0.038 \; (0.052)$         & $-0.005 \; (0.055)$         & $0.053 \; (0.071)$          & $-0.005 \; (0.053)$        \\
Contact            & $-0.039 \; (0.022)^{\cdot}$ & $0.090 \; (0.024)^{***}$    & $0.090 \; (0.029)^{**}$     & $0.032 \; (0.019)^{\cdot}$ \\
Knowledge          & $-0.093 \; (0.036)^{*}$     & $0.003 \; (0.041)$          & $-0.116 \; (0.051)^{*}$     & $0.121 \; (0.037)^{***}$   \\
Numeracy           & $-0.019 \; (0.010)^{\cdot}$ & $0.023 \; (0.011)^{*}$      & $-0.063 \; (0.015)^{***}$   & $0.003 \; (0.010)$         \\
Trust              & $0.623 \; (0.016)^{***}$    & $0.060 \; (0.017)^{***}$    & $-0.014 \; (0.021)$         & $0.019 \; (0.015)$         \\
Age                & $-0.001 \; (0.001)$         & $-0.012 \; (0.001)^{***}$   & $0.008 \; (0.001)^{***}$    & $-0.000 \; (0.001)$        \\
Male               & $-0.021 \; (0.026)$         & $0.030 \; (0.029)$          & $-0.250 \; (0.037)^{***}$   & $-0.080 \; (0.025)^{**}$   \\
GCSE or below      & $0.041 \; (0.042)$          & $-0.082 \; (0.047)^{\cdot}$ & $-0.023 \; (0.061)$         & $0.019 \; (0.044)$         \\
Undergraduate      & $-0.056 \; (0.030)^{\cdot}$ & $0.026 \; (0.033)$          & $-0.032 \; (0.041)$         & $0.014 \; (0.028)$         \\
Postgraduate       & $-0.083 \; (0.040)^{*}$     & $0.048 \; (0.044)$          & $-0.082 \; (0.054)$         & $0.007 \; (0.036)$         \\
Other ethnicity    & $0.062 \; (0.038)$          & $-0.175 \; (0.041)^{***}$   & $0.103 \; (0.051)^{*}$      & $-0.006 \; (0.034)$        \\
Income $>$ 30k     & $0.026 \; (0.025)$          & $0.047 \; (0.027)^{\cdot}$  & $-0.060 \; (0.034)^{\cdot}$ & $-0.005 \; (0.024)$        \\
No party           & $-0.110 \; (0.046)^{*}$     & $-0.122 \; (0.050)^{*}$     & $-0.061 \; (0.064)$         & $0.007 \; (0.047)$         \\
Lib Dem            & $-0.281 \; (0.056)^{***}$   & $-0.168 \; (0.057)^{**}$    & $0.129 \; (0.071)^{\cdot}$  & $0.161 \; (0.048)^{***}$   \\
Labour             & $-0.248 \; (0.043)^{***}$   & $-0.111 \; (0.046)^{*}$     & $0.204 \; (0.057)^{***}$    & $0.095 \; (0.042)^{*}$     \\
Green Party        & $-0.195 \; (0.057)^{***}$   & $-0.222 \; (0.062)^{***}$   & $0.080 \; (0.078)$          & $0.106 \; (0.055)^{\cdot}$ \\
Other party        & $-0.352 \; (0.062)^{***}$   & $-0.192 \; (0.062)^{**}$    & $0.082 \; (0.079)$          & $0.115 \; (0.059)^{*}$     \\
EU: Did not vote   & $-0.067 \; (0.051)$         & $-0.009 \; (0.054)$         & $-0.010 \; (0.068)$         & $0.057 \; (0.051)$         \\
EU: Remain         & $-0.161 \; (0.034)^{***}$   & $-0.037 \; (0.038)$         & $-0.027 \; (0.047)$         & $0.186 \; (0.035)^{***}$   \\
\hline
R$^2$              & $0.509$                     & $0.072$                     & $0.069$                     & $0.050$                    \\
Adj. R$^2$         & $0.505$                     & $0.065$                     & $0.062$                     & $0.043$                    \\
Num. obs.          & $2917$                      & $2917$                      & $2917$                      & $2917$                     \\
\hline
\multicolumn{5}{l}{\scriptsize{$^{***}p<0.001$; $^{**}p<0.01$; $^{*}p<0.05$; $^{\cdot}p<0.1$}}
\end{tabular}
\label{tab8}
\end{center}
\end{table}

\begin{table}[h!]
\caption{Support for restrictions (logistic models with robust standard errors)}
\begin{center}
\begin{tabular}{l c c c}
\hline
 & Main effect & x Knowledge & x Contact \\
\hline
(Intercept)        & $0.470 \; (0.255)^{\cdot}$ & $0.472 \; (0.255)^{\cdot}$ & $0.473 \; (0.255)^{\cdot}$ \\
Cond.: UK-Only     & $0.193 \; (0.109)^{\cdot}$ & $0.193 \; (0.109)^{\cdot}$ & $0.188 \; (0.109)^{\cdot}$ \\
x Knowledge        &                            & $-0.003 \; (0.185)$        &                            \\
x Contact          &                            &                            & $-0.135 \; (0.186)$        \\
Cond.: Comparative & $0.075 \; (0.105)$         & $0.069 \; (0.106)$         & $0.068 \; (0.105)$         \\
x Knowledge        &                            & $-0.094 \; (0.175)$        &                            \\
x Contact          &                            &                            & $-0.201 \; (0.181)$        \\
Contact            & $0.122 \; (0.076)$         & $0.122 \; (0.076)$         & $0.235 \; (0.128)^{\cdot}$ \\
Knowledge          & $0.393 \; (0.076)^{***}$   & $0.429 \; (0.128)^{***}$   & $0.392 \; (0.076)^{***}$   \\
Numeracy           & $0.015 \; (0.036)$         & $0.015 \; (0.036)$         & $0.015 \; (0.036)$         \\
Trust              & $-0.007 \; (0.051)$        & $-0.008 \; (0.052)$        & $-0.007 \; (0.051)$        \\
Age                & $-0.001 \; (0.004)$        & $-0.001 \; (0.004)$        & $-0.001 \; (0.004)$        \\
Male               & $-0.233 \; (0.094)^{*}$    & $-0.235 \; (0.094)^{*}$    & $-0.234 \; (0.094)^{*}$    \\
GCSE or below      & $0.040 \; (0.153)$         & $0.042 \; (0.153)$         & $0.036 \; (0.153)$         \\
Undergraduate      & $0.015 \; (0.105)$         & $0.014 \; (0.105)$         & $0.011 \; (0.105)$         \\
Postgraduate       & $0.001 \; (0.143)$         & $0.001 \; (0.143)$         & $-0.001 \; (0.143)$        \\
Other ethnicity    & $-0.140 \; (0.130)$        & $-0.140 \; (0.130)$        & $-0.139 \; (0.130)$        \\
Income $>$ 30k     & $-0.009 \; (0.089)$        & $-0.008 \; (0.089)$        & $-0.007 \; (0.089)$        \\
No party           & $0.033 \; (0.148)$         & $0.032 \; (0.148)$         & $0.036 \; (0.148)$         \\
Lib Dem            & $0.657 \; (0.201)^{**}$    & $0.659 \; (0.201)^{**}$    & $0.659 \; (0.201)^{**}$    \\
Labour             & $0.319 \; (0.139)^{*}$     & $0.315 \; (0.140)^{*}$     & $0.323 \; (0.139)^{*}$     \\
Green Party        & $0.390 \; (0.197)^{*}$     & $0.391 \; (0.198)^{*}$     & $0.397 \; (0.198)^{*}$     \\
Other party        & $0.547 \; (0.207)^{**}$    & $0.549 \; (0.207)^{**}$    & $0.558 \; (0.207)^{**}$    \\
EU: Did not vote   & $0.205 \; (0.157)$         & $0.204 \; (0.157)$         & $0.207 \; (0.157)$         \\
EU: Remain         & $0.558 \; (0.112)^{***}$   & $0.558 \; (0.113)^{***}$   & $0.558 \; (0.113)^{***}$   \\
\hline
AIC                & $3242.006$                 & $3245.623$                 & $3244.703$                 \\
BIC                & $3367.551$                 & $3383.124$                 & $3382.204$                 \\
Log Likelihood     & $-1600.003$                & $-1599.811$                & $-1599.352$                \\
Deviance           & $3200.006$                 & $3199.623$                 & $3198.703$                 \\
Num. obs.          & $2917$                     & $2917$                     & $2917$                     \\
\hline
\multicolumn{4}{l}{\scriptsize{$^{***}p<0.001$; $^{**}p<0.01$; $^{*}p<0.05$; $^{\cdot}p<0.1$}}
\end{tabular}
\label{tab9}
\end{center}
\end{table}

\begin{table}[h!]
\caption{Main effects without covariates (OLS with robust standard errors)}
\begin{center}
\begin{tabular}{l c c c c}
\hline
 & Gov. Performance & Activity & Concern & Restrictions \\
\hline
(Intercept)                                        & $-0.019 \; (0.030)$ & $0.021 \; (0.024)$  & $-0.014 \; (0.029)$ & $-0.011 \; (0.020)$ \\
Cond.: UK-Only                                     & $0.044 \; (0.043)$  & $-0.028 \; (0.034)$ & $0.021 \; (0.041)$  & $0.038 \; (0.029)$  \\
Cond.: 
                                     Comp. & $0.013 \; (0.042)$  & $-0.035 \; (0.034)$ & $0.022 \; (0.042)$  & $-0.004 \; (0.029)$ \\
\hline
R$^2$                                              & $0.000$             & $0.000$             & $0.000$             & $0.001$             \\
Adj. R$^2$                                         & $-0.000$            & $-0.000$            & $-0.001$            & $0.000$             \\
Num. obs.                                          & $2917$              & $2917$              & $2917$              & $2917$              \\
\hline
\multicolumn{5}{l}{\scriptsize{$^{***}p<0.001$; $^{**}p<0.01$; $^{*}p<0.05$; $^{\cdot}p<0.1$}}
\end{tabular}
\label{tab10}
\end{center}
\end{table}

\begin{figure}[h]
\includegraphics[width=1\linewidth]{covid_appendix_files/figure-latex/fig2-1} \caption{Individual Government Performance Items: Effect compared to Control (95 percent CIs with robust SEs)}\label{fig:fig2}
\end{figure}

\begin{figure}[h]
\includegraphics[width=1\linewidth]{covid_appendix_files/figure-latex/fig3-1} \caption{Condition x Partisanship: Difference in effect for non-Conservatives compared with Conservatives (95 percent CIs with robust SEs)}\label{fig:fig3}
\end{figure}






\end{document}
